Article Text

FRI0539 Reliability of A Semi-Automatic Algorithm in The Detection of Cortical Breaks in Finger Joints Using High Resolution Peripheral Quantitative CT
  1. M. Peters1,
  2. A. Scharmga1,
  3. J. de Jong1,
  4. A. van Tubergen1,
  5. R. Weijers2,
  6. D. Loeffen2,
  7. B. van Rietbergen3,
  8. J. van den Bergh1,
  9. P. Geusens1
  1. 1Rheumatology
  2. 2Radiology, MUMC, Maastricht
  3. 3Biomedical Engineering, TUE, Eindhoven, Netherlands


Background Conventional radiography (CR) is considered the gold standard for diagnosis and monitoring of erosions in rheumatoid arthritis (RA). However, CR is less sensitive in the detection of erosions than high resolution peripheral quantitative CT (HR-pQCT) in hand joints (1). The visual scoring of these erosions is, however, laborious. Therefore an automatic method could be timesaving. We previously developed a semi-automatic algorithm for HR-pQCT that is able to detect cortical breaks. However, the intra-operator reliability of our algorithm was not yet tested on the joint level for different break sizes, and neither on a single break level.

Objectives To test the intra-operator reliability of the algorithm on HR-pQCT images on the joint level and on a single break level for different break sizes.

Methods Nineteen finger joints of ten human female cadaveric index fingers (mean age ± SD; 85.1 ± 9.6 years) with unknown medical history were imaged by HR-pQCT (82μm voxel size). A semi-automatic algorithm was developed for the detection of cortical breaks, taking into account various break sizes. First, the automatically detected contours were manually adapted twice by the same operator several weeks apart. Second, bone within 0.25mm from this contour was selected as the cortical region. Last, different sizes for defining a cortical break (>0.50mm, >0.66mm and >0.82mm) were applied. Breaks that were detected by the algorithm using both contours, and overlapped with at least 30 voxels were counted as matching breaks. The proportion of matching breaks was calculated as a measure of reliability for a single break. Intraclass correlation coefficients (ICCs) were calculated as a measure of reliability on the joint level.

Results The number of breaks depended on the chosen break size, and varied between 2.4 and 14.2 per joint. The locations of the detected breaks from the first and second time contouring generally corresponded well for all break sizes (Fig. 1). The agreement on a single break level was substantial for all break sizes (≥67.7%), with the highest agreement for breaks >0.66mm (80.9%).The agreement on the joint level was high for all break sizes (ICC≥.84, p<0.01) and excellent for breaks >0.66mm (ICC ≥.95, p<0.01) (table 1).

Table 1.

Number of breaks and intra-observer reliability of breaks detected on the HR-pQCT images

Conclusions The intra-operator reliability of the algorithm was excellent at the joint level and very good of the single break detection for all break sizes. Therefore, the use of HR-pQCT in combination with our semi-automatic algorithm is a promising tool for early detection and monitoring of the number of small cortical breaks in finger joints.

  1. Stach et al., A&R.2010 Feb; 62(2):330–339

Disclosure of Interest M. Peters: None declared, A. Scharmga: None declared, J. de Jong: None declared, A. van Tubergen: None declared, R. Weijers: None declared, D. Loeffen: None declared, B. van Rietbergen Consultant for: Scanco Medical AG, J. van den Bergh: None declared, P. Geusens: None declared

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.